Several types of materials are routinely used to fabricate substrates for nitride-based semiconductor structures, which, in turn, can be employed as components of high-performance electronic and optoelectronic devices.
For devices employing GaN or Ga1-xInxN, the most desirable substrate would nominally be a large-area GaN single-crystal wafer. While several methods to grow GaN crystals have been proposed, none of them appears to be commercially feasible to fabricate large-area bulk crystals of GaN.
Sapphire is a popular conventional substrate material, because relatively high-quality, inexpensive sapphire substrates are commercially available. However, sapphire is far from being perfectly suited for GaN epitaxy. Its lattice mismatch to GaN is relatively large (about 16%), it has little distinction between the + and −[0001] directions which can give rise to +/− c-axis domains in epitaxial films of GaN, and its differential thermal expansion may lead to cracking during the post-fabrication cooling process. In spite of those drawbacks, recently, Nichia Ltd of Japan has announced the production of a violet laser with significant commercial possibilities—more than 10,000 hours of operating life—using sapphire substrates. Currently, laser diodes (“LDs”) are relatively expensive. Using sapphire substrates leads to a costly fabrication process because it requires growing buffer layers and using lateral epitaxial overgrowth techniques. Even though Nichia's announcement is very promising, violet lasers grown over sapphire substrates still have shortcomings. For example, heat may build up in these lasers during operation. Sapphire, with its very low thermal conductivity, traps that heat, which may trigger a premature burnout of these devices. To build a more durable blue laser, Nichia and others are investigating other alternatives, such as free-standing GaN substrates. In this technique, the substrate is removed after a thick GaN layer is grown thereover. This method leaves the GaN as the base for building the laser. This base should be better at dissipating heat, in addition to matching the alloy layers disposed thereover. However, this approach further increases process complexity and associated fabrication costs.
Single-crystal substrates of SiC present an attractive alternative to sapphire due to their close lattice match to AlN/GaN in the plane perpendicular to the c-axis (the so-called “c-plane”) and high thermal conductivity. In addition, SiC substrates can be made electrically conducting, which is attractive for some applications, including fabrication of LEDs and LDs. However, wurtzite SiC (matching the wurtzite crystal structure of GaN) is not available and the lattice mismatch along the c-axis between GaN and both 4H and 6H SiC is substantial. In addition, the chemical bonding between the Group IV elements of the SiC and the Group III or Group V elements of the nitrides is likely to create nucleation problems leading to electronic states at the interface.
It is, therefore, desirable to provide alternative substrates for commercial fabrication of nitride-based (e.g., GaN) devices. In particular, the physical and electronic properties of AlN—wide energy bandgap (6.2 electron volts), high breakdown electric field, extremely high thermal conductivity, and low optical density—afford this material a great potential for a wide variety of semiconductor applications as a substrate material. Despite these useful properties of AlN substrates, commercial feasibility of AlN-based semiconductor devices, however, have been limited by the unavailability of large, low-defect single crystals of AlN.
A sublimation-recondensation technique was developed for AlN crystal growth by Slack and McNelly (G. A. Stack and T. McNelly, J. Cryst. Growth 34, 263 (1976) and 42, 560 (1977), hereinafter the “Slack reference,” incorporated herein by reference). In this technique, polycrystalline source material is placed in the hot end of a crucible while the other end is kept cooler. The crystal nucleates in the tip and grows as the crucible is moved through the temperature gradient. This approach demonstrated relatively slow crystal growth of 0.3 mm/hr while the crystal growth chamber was maintained at 1 atm of N2, producing a conical crystal about 12 mm in length and about 4 mm in diameter.
To make AlN substrates and devices built thereon commercially feasible, it would be desirable to increase the growth rate. A number of researchers have examined the possibility of such increase. Many of them, however, relied on rate equations derived by Dryburgh (see “Estimation of maximum growth rate for aluminum nitride crystals by direct sublimation,” J. Crystal Growth 125, 65 (1992)), which appear to overestimate the growth rate of AlN and, in particular, suggest that the maximum growth conditions are near stoichiometric vapor conditions, i.e., the Al and N2 partial pressures should be adjusted so that the Al partial pressure is twice that of the N2. See, for example, U.S. Pat. Nos. 5,858,085; 5,972,109; 6,045,612; 6,048,813; 6,063,185; 6,086,672; and 6,296,956, all to Hunter. In addition, known approaches generally maintain the N2 partial pressure at less than atmospheric pressure.
Most attempts at increasing the growth rate of AlN crystals under such stoichiometric and/or sub-atmospheric pressure conditions have met with limited success. In addition, it appears to be impossible to achieve the growth rate, or the electronics-grade quality Hunter discloses in his patents with the N2 pressure below one atmosphere. Additional research in the field of fabrication of AlN crystals was reported by Segal and his colleagues. (See Segal et al. “On mechanisms of sublimation growth of AlN bulk crystals,” J. Crystal Growth 211, 68 (2000)). This article appears to be the first peer-reviewed publication suggesting that Dryburgh's growth equations are incorrect. Segal and his colleagues, however, teach open growth conditions, allowing the Al vapor to escape. Disadvantageously, it would be difficult to grow large boules of AlN using this approach, because: (i) due to the non-uniform growth across the surface, growth control is difficult; (ii) a large amount of Al would be wasted; (iii) the excess Al in the rest of the furnace would create problems due to of its high reactivity; and (iv) it would be difficult to maintain a high temperature differential between the source and growing crystal surface.
Critical to the ability to made commercially practical nitride-based semiconductor devices, is achieving low levels of defect densities throughout the semiconductor structure. With conventional nitride semiconductor growth techniques on commonly available foreign substrates or high-defect nitride substrates, a GaN buffer layer is grown thick to achieve planar growth fronts and to relax the GaN epitaxial layers prior to formation of the active region of the device, i.e. the GaN/AlGaN heterostructure. This approach will result in epitaxial layers with defect densities in the 108 cm−2 to 1010 cm−2 range.
Other researchers have attempted to reduce dislocation densities by epitaxial lateral overgrowth, novel nucleation schemes to initiate growth on the foreign substrates, or by adding complex structure such as superlattices into the epitaxial profile. For example, Sumitomo Electric has reported local regions of 105 dislocations per cm2 although the average dislocation density in their GaN substrates exceeds 106 cm−2. Also, researchers from Tokyo University of Agriculture and Technology and TDI reported the defect densities of 107 cm−2. Work by others on AlN templates can be expected to have a large defect density due to the initial growth on a foreign substrate, even though the original substrate might be removed to obtain a freestanding AlN wafer.
Thus, present state-of-the-art technology employs foreign substrates that have large thermal and lattice differences relative to AlN, AlGaN, and GaN epitaxial layers. This results in defect densities ranging from 108 cm−2 to 1010 cm−2, which makes certain devices—especially devices employing AlGaN layers with high Al content—unrealizable.
A need therefore exists for large AlN substrates suitable for fabricating semiconductor devices thereon and commercially-feasible methods for manufacturing these substrates that address the aforementioned drawbacks.